Interesting I may have been comparing parallel bzip2 with single threaded xz or perhaps it was just the data I was using for a test.

I'll have to look at it again, thanks!